R plotly version 4.5.2 scatterplot legend bubble size settings

Question

I am using plotly 4.5.2 in R. I have created a scatterplot which is sized on a variable, the issue is these sizes are also reflected in the legend which makes them hard to read.

I want my graph to remain the same, with the only exception being the size of the bubbles in the legend. These bubbles can either be set to all be the same size or scaled to a smaller size. Importantly, the sizes in the graph must remain the same.

Please find reproducible code here:

library(plotly)

data <- data.frame(name = c('test1', 'test2', 'test3', 'test4'),
                      x = c(1, 15, 90, 45),
                      y = c(9, 43, 43, 53),
                      size = c(10000, 50000, 90000, 3000),
                      colour = c("rgba(230, 42, 56, 0.3)", "rgba(76, 175, 80, 0.3)",
                                 "rgba(32, 169, 242, 0.3)", "rgba(255, 193, 7, 0.3)")
                      )

plot <- plot_ly(data = data) %>% 
  add_trace(x = ~x,
            y = ~y,
            mode = 'markers',
            type = 'scatter',
            color = ~name,
            marker = list(
              color = ~colour,               
              opacity = 1,
              showlegend=T),
            size = ~size)

Thank you

I'm not completely sure, but I think it is (so far) not possible as you can read [here](http://stackoverflow.com/questions/37245004/change-legend-size-in-plotly-chart) or in the [legend reference](https://plot.ly/r/reference/#layout-legend). There is no argument to reset the size of the legend entries. — Roman, Oct 18 '16 at 11:18

score 5 · Accepted Answer · answered Oct 21 '16 at 01:30

I found a hack to get the desired output, i'm posting it here for the benefit of others.

library(plotly)

data <- data.frame(name = c('test1', 'test2', 'test3', 'test4'),
                      x = c(1, 15, 90, 45),
                      y = c(9, 43, 43, 53),
                      size = c(10000, 50000, 90000, 3000),
                      colour = c("rgba(230, 42, 56, 0.3)", "rgba(76, 175, 80, 0.3)",
                                 "rgba(32, 169, 242, 0.3)", "rgba(255, 193, 7, 0.3)")
                      )


#Ranges
xmin <- - 0.2 * max(data[['x']])
xmax <- 1.8 * max(data[['x']])
ymin <- - 0.2 * max(data[['y']])
ymax <- 1.8 * max(data[['y']])


# Sum of the size variable
sum_size <- sum(data[['size']], na.rm = TRUE)

# Decimal size
data$size <- (data[['size']]/sum_size)

# Adjust for the smallest 
data <- data %>% mutate(size = ifelse(size < 0.05, 0.05, size))

#Size Vector
size <- data$size * 100

# not used atm
min_size <- min(data$size, na.rm = TRUE)
max_size <- max(data$size, na.rm = TRUE)


# Number of unique groups
num_bubbles <- length(unique(data[['name']])) 


# Artifical data used to resolve legend sizes
data2 <- data
data2$size <- min_size
data2[['x']] <- -2 * max(-xmin,-ymin)
data2[['y']] <- -2 * max(-xmin,-ymin)

# Bind the artifial data, plotly will only plot the original and this fixes the legend size issue
data <- rbind(data, data2)

plot <- plot_ly(data = data) %>% 
  add_trace(x = data[['x']],
            y = data[['y']],
            mode = 'markers',
            type = 'scatter',
            color = data[['name']], 
            marker = list(size = 10,
                          opacity = 1,sizemin=10,sizemax =100,sizeref = 100,
                          line = list(width = 2)),size = 30,showlegend=T,
            hoverinfo = "text") %>% 
  add_trace( x = -2 * max(-xmin,-ymin) , y = -2 * max(-xmin,-ymin), type = "scatter", mode = "markers", 
             color= data[['name']], showlegend=F) %>% config(modeBarButtonsToRemove = list("sendDataToCloud","pan2d","select2d","lasso2d","zoomIn2d","zoomOut2d","autoScale2d","resetScale2d","hoverClosestCartesian","hoverCompareCartesian"), displaylogo = FALSE, doubleClick = "reset")  



plot <- layout(plot,
               title = NULL,

               xaxis = list(           

                 title = 'x',
                 range = c(xmin,xmax),
                 showgrid = F     
               ),
               yaxis = list(         
                 title = 'y',

                 range = c(ymin,ymax)
               ))

plot <- plotly_build(plot)

for(i in seq(1,num_bubbles))
{
  plot$x$data[[i]]$marker$size <- c(size[i]*10000,min(size)*10000)


}

I have been trying this solution, but it becomes a problem when your data set is bigger sets per trace. I found a solution for it that does not require the negative data points to work. It can't work for the variable size of markers like you have though. — Mark, Jan 22 '18 at 21:33
Basically what happens with this hack is that plotly makes points 1:i in the plot the size of the values 1:i you send to plot$x$data.... and for the legend it takes an average of all the values you send. The size in the legend becomes limited however to the line height of the legend, as is the case here in your example. You can't make them bigger than this (even though originally they were much larger ), but you can shrink them down to 1 pixel. In my example I can manage any marker size in plot (uniform size) and any size in the legend) thanks to your code example and some testing — Mark, Jan 22 '18 at 21:38

score 1 · Answer 2 · answered Jan 19 '18 at 12:09

1

Not sure how to incorporate this into plotly itself, but if you encounter such problem in Shiny app try using this in css file or use inlineCSS in UI:

.legendpoints path.scatterpts {
  d: path('M 7.5 0 A 7.5 7.5 0 1 1 0 -7.5 A 7.5 7.5 0 0 1 7.5 0 Z');
}

Replace each '7.5' with the half of px size for the legend entry (so here it will be 15px)

How it looks in plotly in app

answered Jan 19 '18 at 12:09

M.Dubel

126
1
9

Adding the CSS above worked perfectly when building a GitBook with bookdown. Thank you! – Daniel L Mar 18 '22 at 14:27

score -2 · Answer 3 · answered Oct 19 '16 at 08:04

-2

You can use ggplot and the ggplotly() function:

library(ggplot2)
p <- ggplot(data, aes(x=x, y=y, size=size, color=name)) + 
           geom_point() + 
           theme_minimal() + 
           theme(legend.title=element_blank())
ggplotly(p)

But there is the same problem with plotly_4.5.2 and the legend functionalities. Both ways to remove the size from the legend are working in ggplot2_2.1.0 but not in plotly using the transformation function ggplotly().

p <- p + guides(size=FALSE)
p <- p + scale_size_continuous(guide=FALSE)
ggplotly(p)

answered Oct 19 '16 at 08:04

Roman

17,008
3
36
49

Thanks Jimbou, I managed to reach this far as well. As mentioned, its paramount that the scale of the bubbles is the same. I tried to change the sizeref, sizemin and sizemax values but cannot get the desired graph. – ZeroStack Oct 20 '16 at 01:23
@ZeroStack "the scale of the bubbles is the same" compared to what? ggplot plots the bubbles size compared to the values in the size vector. – Roman Oct 20 '16 at 09:59
@Jimbou - he means that the size of the bubbles in the legend should be the same, as in his "hack". – Chechy Levas Aug 24 '17 at 08:52

R plotly version 4.5.2 scatterplot legend bubble size settings

3 Answers3

Linked